Timing is everything: using AI disease forecasting to protect your vineyard
A real-world walkthrough of how databaum's incubation model tells you not just whether to spray — but precisely when.
For downy mildew, the difference between a controlled situation and a serious outbreak often comes down to a few days. Spray too early and your fungicide loses efficacy before peak infection pressure arrives. Spray too late and oil spot lesions are already forming on the leaf. The challenge is that the ideal intervention window doesn't announce itself — it has to be calculated continuously as weather evolves.
This is exactly what databaum's AI disease model is designed to do. Rather than offering a generic "high risk" alert, it tracks incubation progress in near real-time so you can make spray decisions with precision. Below, we walk through a real example from our weekly forecast dashboard.
Reading the weekly forecast view
The dashboard shows three rows of information for each date:
OSL (Oil Spot Lesion — when visible symptoms may appear on the leaf),
80% INC (the ideal timing for a systemic fungicide), and
INF (when an infection event began). Below those rows, incubation progress curves trace the development of active infections over time.
Each date in the calendar shows a percentage indicating how far the current infection has progressed. A reading of 44% means the infection is nearly halfway through its incubation cycle. Once it approaches 80%, the treatment window is open.
A real example: the May 6 infection
The screenshot below shows the databaum dashboard for May 10–23, 2026. Two active infections are visible — one that started before May 10 (the red curve, nearing completion) and a newer one that began around May 15 (the orange curve, just rising).
[Screenshot: databaum AI weekly disease forecast, May 10–23, 2026. The calendar shows the May 6 infection at 100% completion dated May 17, and a second infection at 44% on May 15.]
Let's trace what the model was signalling for the infection that started on May 6.
The infection began on May 6, with the INF row lighting up on that date. Over the following days, incubation progressed steadily as weather conditions drove development. By May 15–16, the 80% INC marker turned orange — the model's primary signal that a systemic fungicide should be applied. The ideal application window was May 15 or 16, one day before the 80% threshold was fully reached. At this point, systemic products can still move through plant tissue and interrupt sporulation before it takes hold.
By May 17, incubation was complete at 100%. The OSL row turned orange on that date, indicating that oil spot lesions could become visible on the leaf surface. If no treatment had been applied, the window for effective systemic intervention had passed.
The ideal timing for a systemic fungicide is when the incubation progress curve is approaching 80% — not after it crosses it. On the May 6 infection, that meant acting on May 15 or 16.
A second infection arrives on May 15
Looking at the same dashboard, the INF row shows a new infection event starting on May 15 — visible as the orange-bordered tile on that date, with the second incubation curve beginning to rise.
This creates an important decision point, depending on what was sprayed previously.
If you applied a systemic fungicide on May 15 or 16 to address the first infection, that same treatment very likely covers the new infection as well. Systemic products act on early-stage infections already inside plant tissue, so one well-timed application can address both infections simultaneously.
If your previous spray was a contact fungicide only, the situation is different. Contact products do not penetrate plant tissue and cannot reach an infection that has already begun inside the leaf. In that case, a contact application on May 14 or 15 could have protected exposed surface tissue — or a follow-up systemic treatment should be planned before the new infection's own 80% threshold is reached.
The takeaway: when a new INF event appears shortly before or during an active treatment window, one well-timed systemic application can clean up both infections at once. But if contact-only products have been used, the new infection requires its own dedicated response.
The recommended workflow
Once you have an active infection on the calendar, the process is straightforward:
- Monitor the daily AI pressure view for persistent colour signals across consecutive days
- Check INF dates on the calendar — note when new infections have started
- Watch the incubation progress curve — plan your spray as it approaches 80%
- Check the spraying window for rainfall risk, humidity, and drying conditions
- Apply the treatment and record it — the shield indicator becomes active
- Monitor the shield — once it expires, return to step 1
As long as the shield is active, the next spray can generally be skipped. Once it disappears, a new infection cycle may already be underway.
A few practical notes
On organic products: Contact and organic PPPs have less precisely defined protection windows and may need more frequent application. Treat the 80% INC signal as a hard deadline rather than a flexible range.
On Piwi varieties: Continue following cantonal service and Agroscope guidance for preventive sprays around flowering. The model supports, but does not replace, variety-specific protocols.
On forecast horizon: The platform shows a 4-day forecast only. Weather predictions beyond four days introduce enough uncertainty that disease development forecasts become unreliable. A shorter, more accurate forecast is more useful than a longer, uncertain one.
The AI model is a decision-support tool, not an automatic spray instruction. The most effective growers use it to avoid unnecessary treatments when pressure is genuinely low — and to ensure they never miss a critical window when it matters most.
At databaum, our goal is to make disease management more precise, transparent, and practical for real vineyard operations.